Random sampling
-Rcpp::IntegerVector dqrng::dqsample_int(int n, int size, bool replace = false,
- Rcpp::Nullable<Rcpp::NumericVector> probs = R_NilValue,
- int offset = 0)
-Rcpp::NumericVector dqrng::dqsample_num(double n, double size, bool replace = false,
- Rcpp::Nullable<Rcpp::NumericVector> probs = R_NilValue,
- int offset = 0)Rcpp::IntegerVector dqrng::dqsample_int(int n, int size, bool replace = false,
+ Rcpp::Nullable<Rcpp::NumericVector> probs = R_NilValue,
+ int offset = 0)
+Rcpp::NumericVector dqrng::dqsample_num(double n, double size, bool replace = false,
+ Rcpp::Nullable<Rcpp::NumericVector> probs = R_NilValue,
+ int offset = 0)n-
@@ -230,6 +241,92 @@
Random sampling
The two functions are used for “normal” and “long-vector” support in R.
+
+
+
+Getting and setting the RNG state +
+std::vector<std::string> dqrng::dqrng_get_state()
+void dqrng::dqrng_set_state(std::vector<std::string> state)-
+
state
+-
+a
std::vector<std::string>as produced by +dqrng_get_state()+
+
+
+
@@ -250,7 +347,7 @@ Accessing the global RNG +
+ +Direct usage of this method is discouraged. The
+prefered way of accesing the global RNG is to instantiate
+dqrng::random_64bit_accessor within your function. Note
+that you MUST NOT delete the global RNG. Using
+dqrng::random_64bit_accessor makes this impossible. In
+addition, you SHOULD NOT store a reference to the RNG permanently,
+because it can be invalidated by calls to dqRNGkind.
+Therefore, instances of dqrng::random_64bit_accessor SHOULD
+be stored as (non-static) variables in functions.
Note that dqrng::random_64bit_accessor supports UniformRandomBitGenerator
+and can therefore be used together with any C++11 distribtion function.
+In addition, the follwing functions are supported, since they are
+inherited from the abstract parent class
+random_64bit_generator:
// clone RNG and seelct a different stream
+std::unique_ptr<random_64bit_generator> clone(uint64_t stream)
+// uniform doubles in [0,1) and double-int-pairs
+double uniform01()
+std::pair<double, int> generate_double_8bit_pair()
+// uniform integers in a range
+uint32_t operator() (uint32_t range)
+uint64_t operator() (uint64_t range)
+ResultType variate<ResultType, DistTmpl>(param1, ... paramN)
+generate<DistTmpl>(container, param1, ... paramN)
+generate<DistTmpl>(start, end, param1, ... paramN)
+Dist::result_type variate<Dist>(param1, ... paramN)
+generate<Dist>(container, param1, ... paramN)
+generate<Dist>(start, end, param1, ... paramN)-
+
stream
+- +RNG stream to use for the cloned RNG + +
range
+-
+Integers are generated in closed interval
[0, range]+
+ ResultType
+-
+Expected result from the distribution template
DistTmpl+
+ DistTmpl
+-
+Distribution template like
std::uniform_distribution. +DistTmpl<ResultType>defines the full distribution. +
+ Dist
+-
+Full distribution like
+
std::uniform_distribution<double>or +dqrng::normal_distriubtion. +
+ param1, ... paramN
+- +Necessary parameters to initialize the distribution. + +
container
+-
+A container that is to be filled with variates from the distribution
+function. Needs to support
std::beginand +std::end. +
+ start, end
+- +Forward iterators pointing to start and end of a range to be filled with +variates from the distribution function. + +
Random sampling
-Site built with pkgdown 2.0.7.
+Site built with pkgdown 2.0.8.
diff --git a/articles/dqrng.html b/articles/dqrng.html
index 69a374c..1c4976c 100644
--- a/articles/dqrng.html
+++ b/articles/dqrng.html
@@ -6,7 +6,7 @@
Fast Pseudo Random Number Generators for R • dqrng
-
+
@@ -33,7 +33,7 @@
@@ -91,7 +91,7 @@ Fast Pseudo Random Number Generators for R
Ralf
Stubner
- 2024-04-09
+ 2024-04-15
Source: vignettes/dqrng.Rmd
dqrng.Rmd
@@ -166,7 +166,7 @@ Usage from Rsystem.time(cat("pi ~= ", piR(N), "\n"))
#> pi ~= 3.140899
#> user system elapsed
-#> 0.364 0.044 0.408
+#> 0.370 0.029 0.399
Using dqrng is about three times faster:
library(dqrng)
@@ -174,46 +174,46 @@ Usage from Rsystem.time(cat("pi ~= ", piR(N, rng = dqrng::dqrunif), "\n"))
#> pi ~= 3.141457
#> user system elapsed
-#> 0.222 0.041 0.262
+#> 0.170 0.035 0.206
Since the calculations add a constant off-set, the speed-up for the
RNGs alone has to be even greater:
system.time(stats::runif(N))
#> user system elapsed
-#> 0.078 0.013 0.090
+#> 0.087 0.003 0.090
system.time(dqrng::dqrunif(N))
#> user system elapsed
-#> 0.049 0.008 0.056
+#> 0.023 0.007 0.031
Similar for the exponential distribution:
system.time(stats::rexp(N))
#> user system elapsed
-#> 0.338 0.004 0.342
+#> 0.342 0.005 0.346
system.time(dqrng::dqrexp(N))
#> user system elapsed
-#> 0.063 0.008 0.071
+#> 0.047 0.000 0.047
And for the normal distribution:
system.time(stats::rnorm(N))
#> user system elapsed
-#> 0.367 0.000 0.367
+#> 0.366 0.000 0.366
system.time(dqrng::dqrnorm(N))
#> user system elapsed
-#> 0.086 0.008 0.093
+#> 0.066 0.003 0.070
As well as for sampling with and without replacement:
system.time(for (i in 1:100) sample.int(N, N/100, replace = TRUE))
#> user system elapsed
-#> 0.573 0.004 0.577
+#> 0.576 0.000 0.576
system.time(for (i in 1:100) dqsample.int(N, N/100, replace = TRUE))
#> user system elapsed
-#> 0.031 0.016 0.047
+#> 0.031 0.017 0.047
system.time(for (i in 1:100) sample.int(N, N/100))
#> user system elapsed
-#> 1.533 0.304 1.837
+#> 1.430 0.293 1.722
system.time(for (i in 1:100) dqsample.int(N, N/100))
#> user system elapsed
-#> 0.061 0.008 0.069
+#> 0.051 0.007 0.059
It is also possible to register the supplied generators as
user-supplied RNGs. This way set.seed() and
dqset.seed() influence both (dq)runif and
@@ -373,15 +373,14 @@
Using the header only library// [[Rcpp::export]]
Rcpp::NumericVector splitmix_rnorm(const int n, const double mean = 0.0, const double sd = 1.0) {
auto rng = dqrng::generator<SplitMix>(42);
- dqrng::normal_distribution dist(mean, sd);
- Rcpp::NumericVector result(n);
- std::generate(result.begin(), result.end(), [&dist, &rng](){return dist(*rng);});
- return result;
-}
-/*** R
-splitmix_rnorm(10)
-system.time(splitmix_rnorm(1e7))
-*/
+ Rcpp::NumericVector result(n);
+ rng->generate<dqrng::normal_distribution>(result, mean, sd);
+ return result;
+}
+/*** R
+splitmix_rnorm(10)
+system.time(splitmix_rnorm(1e7))
+*/
Since SplitMix is a very fast RNG, the speed of this function is
comparable to dqrnorm. Generally speaking you can use any
C++11 compliant RNG with 64 bit output size. For example, here the 64
@@ -395,16 +394,15 @@
Using the header only library// [[Rcpp::export]]
Rcpp::NumericVector threefry_rnorm(const int n, const double mean = 0.0, const double sd = 1.0) {
auto rng = dqrng::generator<sitmo::threefry_13_64>(42);
- dqrng::normal_distribution dist(mean, sd);
- Rcpp::NumericVector result(n);
- std::generate(result.begin(), result.end(), [&dist, &rng](){return dist(*rng);});
- return result;
-}
-
-/*** R
-threefry_rnorm(10)
-system.time(threefry_rnorm(1e7))
-*/
+ Rcpp::NumericVector result(n);
+ rng->generate<dqrng::normal_distribution>(result, mean, sd);
+ return result;
+}
+
+/*** R
+threefry_rnorm(10)
+system.time(threefry_rnorm(1e7))
+*/
Note that for the (recommended) Threefry engine with 20 rounds some
additional integration is provided in the dqrng_threefry.h
header file.
@@ -422,16 +420,15 @@ Using the header only library// [[Rcpp::export]]
Rcpp::NumericVector std_rnorm(const int n, const double mean = 0.0, const double sd = 1.0) {
auto rng = dqrng::generator<dqrng::xoroshiro128plus>(42);
- std::normal_distribution<double> dist(mean, sd);
- Rcpp::NumericVector result(n);
- std::generate(result.begin(), result.end(), [&dist, &rng](){return dist(*rng);});
- return result;
-}
-
-/*** R
-std_rnorm(10)
-system.time(std_rnorm(1e7))
-*/
+ Rcpp::NumericVector result(n);
+ rng->generate<std::normal_distribution>(result, mean, sd);
+ return result;
+}
+
+/*** R
+std_rnorm(10)
+system.time(std_rnorm(1e7))
+*/
Typically this is not as fast as dqrnorm, but the
technique is useful to support distributions not (yet) included in
dqrng. Note however, that the algorithms used for the distributions from
@@ -560,7 +557,7 @@
Accessing the global RNG
-Site built with pkgdown 2.0.7.
+Site built with pkgdown 2.0.8.
diff --git a/articles/index.html b/articles/index.html
index b8d0e76..26dfae2 100644
--- a/articles/index.html
+++ b/articles/index.html
@@ -1,5 +1,5 @@
-Articles • dqrng Articles • dqrng
@@ -17,7 +17,7 @@
@@ -87,7 +87,7 @@ All vignettes
diff --git a/articles/parallel.html b/articles/parallel.html
index 4111e95..795fb52 100644
--- a/articles/parallel.html
+++ b/articles/parallel.html
@@ -6,7 +6,7 @@
Parallel RNG usage • dqrng
-
+
@@ -33,7 +33,7 @@
@@ -91,7 +91,7 @@ Parallel RNG usage
Ralf
Stubner
- 2024-04-09
+ 2024-04-15
Source: vignettes/parallel.Rmd
parallel.Rmd
@@ -204,7 +204,7 @@ Xo(ro)shiro: jump ahead with OpenMP// [[Rcpp::export]]
std::vector<double> parallel_random_sum(int n, int m, int ncores) {
dqrng::uniform_distribution dist(0.0, 1.0); // Uniform distribution [0,1)
- dqrng::xoshiro256plusplus rng(42); // properly seeded rng
+ dqrng::xoshiro256plusplus rng(42); // properly seeded rng
std::vector<double> res(m);
// ok to use rng here
@@ -239,6 +239,71 @@ Xo(ro)shiro: jump ahead with OpenMP
+In the previous example it was necessary to create our own RNG
+object. It would be nice if it were possible to use dqrng’s global RNG
+instead. To showcase this, we implement the same procedure as before,
+but this time make use of dqrng::random_64bit_accessor to
+access the global RNG and its clone(stream) method that
+creates a cloned version with a different stream:
+#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng.h>
+#include <dqrng_distribution.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::plugins(openmp)]]
+#include <omp.h>
+
+// [[Rcpp::export]]
+std::vector<double> random_sum(int n, int m) {
+ dqrng::uniform_distribution dist(0.0, 1.0); // Uniform distribution [0,1)
+ auto rng = dqrng::random_64bit_accessor{}; // properly seeded rng
+ std::vector<double> res(m);
+ for (int i = 0; i < m; ++i) {
+ double lres(0);
+ for (int j = 0; j < n; ++j) {
+ lres += dist(rng);
+ }
+ res[i] = lres / n;
+ }
+ return res;
+}
+
+// [[Rcpp::export]]
+std::vector<double> parallel_random_sum(int n, int m, int ncores) {
+ dqrng::uniform_distribution dist(0.0, 1.0); // Uniform distribution [0,1)
+ auto rng = dqrng::random_64bit_accessor{}; // properly seeded rng
+ std::vector<double> res(m);
+ // ok to use rng here
+
+#pragma omp parallel num_threads(ncores)
+{
+ // make thread local copy of rng and advance it by 1 ... ncores jumps
+ auto lrng = rng.clone(omp_get_thread_num() + 1);
+
+#pragma omp for
+ for (int i = 0; i < m; ++i) {
+ double lres(0);
+ for (int j = 0; j < n; ++j) {
+ lres += dist(*lrng);
+ }
+ res[i] = lres / n;
+ }
+}
+// ok to use rng here
+return res;
+}
+
+/*** R
+bm <- bench::mark(
+ parallel_random_sum(1e7, 8, 4),
+ random_sum(1e7, 8),
+ check = FALSE
+)
+bm[,1:4]
+*/
+Notes that the thread local RNG uses std::unique_ptr
+instead of Rcpp::Xptr since it is used in parallel
+code.
PCG: multiple streams with RcppParallel
@@ -260,66 +325,110 @@ PCG: multiple streams with RcppP
wrap it as dqrng::generator. The resulting correlation
matrix should be close to the identity matrix if the different streams
are independent:
-#include <Rcpp.h>
-// [[Rcpp::depends(dqrng, BH)]]
-#include <pcg_random.hpp>
-#include <dqrng_sample.h>
-// [[Rcpp::plugins(cpp11)]]
-// [[Rcpp::depends(RcppParallel)]]
-#include <RcppParallel.h>
-
-struct RandomFill : public RcppParallel::Worker {
- RcppParallel::RMatrix<int> output;
- uint64_t seed;
-
- RandomFill(Rcpp::IntegerMatrix output, const uint64_t seed) : output(output), seed(seed) {};
-
- void operator()(std::size_t begin, std::size_t end) {
- auto rng = dqrng::generator<pcg64>(seed, end);
- for (std::size_t col = begin; col < end; ++col) {
- auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*rng, 100000, output.nrow(), true);
- RcppParallel::RMatrix<int>::Column column = output.column(col);
- std::copy(sampled.begin(), sampled.end(), column.begin());
- }
- }
-};
-
-// [[Rcpp::export]]
-Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
- Rcpp::IntegerMatrix res(n, m);
- RandomFill randomFill(res, 42);
- RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
- return res;
-}
-
-/*** R
-res <- parallel_random_matrix(1e6, 8, 4)
-head(res)
-symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
- symbols = c(" ", "?", "!", "1"),
- abbr.colnames = FALSE, corr = TRUE)
-*/
+#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng_generator.h>
+#include <dqrng_sample.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+
+struct RandomFill : public RcppParallel::Worker {
+ RcppParallel::RMatrix<int> output;
+ uint64_t seed;
+
+ RandomFill(Rcpp::IntegerMatrix output, const uint64_t seed) : output(output), seed(seed) {};
+
+ void operator()(std::size_t begin, std::size_t end) {
+ auto rng = dqrng::generator<pcg64>(seed, end);
+ for (std::size_t col = begin; col < end; ++col) {
+ auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*rng, 100000, output.nrow(), true);
+ RcppParallel::RMatrix<int>::Column column = output.column(col);
+ std::copy(sampled.begin(), sampled.end(), column.begin());
+ }
+ }
+};
+
+// [[Rcpp::export]]
+Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
+ Rcpp::IntegerMatrix res(n, m);
+ RandomFill randomFill(res, 42);
+ RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
+ return res;
+}
+
+/*** R
+res <- parallel_random_matrix(1e6, 8, 4)
+head(res)
+symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
+ symbols = c(" ", "?", "!", "1"),
+ abbr.colnames = FALSE, corr = TRUE)
+*/
Head of the random matrix:
[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8]
-[1,] 67984 16279 69262 7126 21441 37720 51107 51045
-[2,] 69310 21713 82885 81157 54051 5261 91165 17833
-[3,] 76742 31232 78953 4626 94939 29416 85652 78296
-[4,] 76349 47427 1770 37957 33888 59134 94591 65793
-[5,] 85008 89224 43493 7925 60866 2464 14080 10763
-[6,] 38017 88509 51195 73086 1883 68193 75259 62216
+[1,] 16144 89512 16144 53103 16144 41760 16144 99788
+[2,] 21200 86552 21199 38327 21198 90293 21198 45713
+[3,] 124 60045 97553 79574 99977 84423 87126 35715
+[4,] 6590 93044 18536 98945 66969 53338 38426 53560
+[5,] 80618 42051 15815 79380 71590 98006 87220 63036
+[6,] 49954 90168 12614 32488 54145 68029 34733 48989
Correlation matrix:
[1,] 1
[2,] 1
-[3,] ? 1
-[4,] ? 1
-[5,] 1
-[6,] ? ? ? 1
-[7,] ? 1
-[8,] ? 1
+[3,] ? 1
+[4,] ? 1
+[5,] ? ? 1
+[6,] ? 1
+[7,] ? 1
+[8,] ? ? 1
attr(,"legend")
[1] 0 ‘ ’ 0.001 ‘?’ 0.003 ‘!’ 0.999 ‘1’ 1
So as expected the correlation matrix is almost equal to the identity
matrix.
+As in the previous section, we can again switch from seeding our own
+RNG to starting from the global RNG and selecting separate substreams
+from there. RNG selection and seeding can then be handled from R:
+#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng.h>
+#include <dqrng_sample.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+
+struct RandomFill : public RcppParallel::Worker {
+ RcppParallel::RMatrix<int> output;
+ dqrng::random_64bit_accessor rng{};
+
+ RandomFill(Rcpp::IntegerMatrix output) : output(output) {};
+
+ void operator()(std::size_t begin, std::size_t end) {
+ auto lrng = rng.clone(end);
+ for (std::size_t col = begin; col < end; ++col) {
+ auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*lrng, 100000, output.nrow(), true);
+ RcppParallel::RMatrix<int>::Column column = output.column(col);
+ std::copy(sampled.begin(), sampled.end(), column.begin());
+ }
+ }
+};
+
+// [[Rcpp::export]]
+Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
+ Rcpp::IntegerMatrix res(n, m);
+ RandomFill randomFill(res);
+ RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
+ return res;
+}
+
+/*** R
+dqRNGkind("pcg64")
+dqset.seed(42)
+res <- parallel_random_matrix(1e6, 8, 4)
+head(res)
+symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
+ symbols = c(" ", "?", "!", "1"),
+ abbr.colnames = FALSE, corr = TRUE)
+*/
@@ -346,7 +455,7 @@ PCG: multiple streams with RcppP
diff --git a/articles/sample.html b/articles/sample.html
index aa28311..25dbcc8 100644
--- a/articles/sample.html
+++ b/articles/sample.html
@@ -6,7 +6,7 @@
Fast sampling methods • dqrng
-
+
@@ -33,7 +33,7 @@
@@ -91,7 +91,7 @@ Fast sampling methods
Ralf
Stubner
- 2024-04-09
+ 2024-04-15
Source: vignettes/sample.Rmd
sample.Rmd
@@ -418,7 +418,7 @@ Technicalities
-Site built with pkgdown 2.0.7.
+Site built with pkgdown 2.0.8.
diff --git a/authors.html b/authors.html
index 5b680d0..41e3ef5 100644
--- a/authors.html
+++ b/authors.html
@@ -1,5 +1,5 @@
-Authors and Citation • dqrng Authors and Citation • dqrng
@@ -17,7 +17,7 @@
@@ -63,7 +63,7 @@
diff --git a/index.html b/index.html
index 3a92a7e..ed1225b 100644
--- a/index.html
+++ b/index.html
@@ -6,26 +6,13 @@
Fast Pseudo Random Number Generators • dqrng
-
+
-
+
Changelog • dqrng Changelog • dqrng
@@ -17,7 +17,7 @@
@@ -163,7 +163,7 @@ dqrng 0.0.1
diff --git a/pkgdown.yml b/pkgdown.yml
index 6e46253..eee6352 100644
--- a/pkgdown.yml
+++ b/pkgdown.yml
@@ -1,10 +1,10 @@
pandoc: 3.1.11
-pkgdown: 2.0.7
+pkgdown: 2.0.8
pkgdown_sha: ~
articles:
cpp-api: cpp-api.html
dqrng: dqrng.html
parallel: parallel.html
sample: sample.html
-last_built: 2024-04-09T20:28Z
+last_built: 2024-04-15T09:24Z
diff --git a/reference/dqrmv-1.png b/reference/dqrmv-1.png
index 64bdb36..2484a48 100644
Binary files a/reference/dqrmv-1.png and b/reference/dqrmv-1.png differ
diff --git a/reference/dqrmv.html b/reference/dqrmv.html
index a919916..709ac47 100644
--- a/reference/dqrmv.html
+++ b/reference/dqrmv.html
@@ -1,5 +1,5 @@
-Multivariate Distributions — dqrmvnorm • dqrng Multivariate Distributions — dqrmvnorm • dqrng
@@ -17,7 +17,7 @@
@@ -101,11 +101,11 @@ Examples
sigma <- matrix(c(4,2,2,3), ncol=2)
x <- dqrmvnorm(n=500, mean=c(1,2), sigma=sigma)
colMeans(x)
-#> [1] 0.9358251 2.0251444
+#> [1] 0.8614361 2.0263967
var(x)
#> [,1] [,2]
-#> [1,] 3.909685 1.884089
-#> [2,] 1.884089 2.976024
+#> [1,] 3.930309 1.917535
+#> [2,] 1.917535 2.881480
plot(x)
@@ -122,7 +122,7 @@ Examples
diff --git a/reference/dqrng-functions.html b/reference/dqrng-functions.html
index 7d471a1..29ebeff 100644
--- a/reference/dqrng-functions.html
+++ b/reference/dqrng-functions.html
@@ -1,5 +1,5 @@
-R interface — dqRNGkind • dqrng R interface — dqRNGkind • dqrng
dqrng
- 0.3.2.2
+ 0.3.2.4
@@ -250,7 +250,7 @@ Examples
diff --git a/reference/dqrng-package.html b/reference/dqrng-package.html
index e12ea13..99547af 100644
--- a/reference/dqrng-package.html
+++ b/reference/dqrng-package.html
@@ -1,5 +1,5 @@
-dqrng: Fast Pseudo Random Number Generators — dqrng-package • dqrng dqrng: Fast Pseudo Random Number Generators — dqrng-package • dqrng Unbiased Random Samples and Permutations — dqsample • dqrng Unbiased Random Samples and Permutations — dqsample • dqrng
@@ -17,7 +17,7 @@
@@ -116,7 +116,7 @@ See also
diff --git a/reference/generateSeedVectors.html b/reference/generateSeedVectors.html
index ebd8b72..b58534a 100644
--- a/reference/generateSeedVectors.html
+++ b/reference/generateSeedVectors.html
@@ -1,5 +1,5 @@
-Generate seed as a integer vector — generateSeedVectors • dqrng Generate seed as a integer vector — generateSeedVectors • dqrng
@@ -17,7 +17,7 @@
@@ -179,7 +179,7 @@ Examples
diff --git a/reference/index.html b/reference/index.html
index 6aedbb9..1d9898d 100644
--- a/reference/index.html
+++ b/reference/index.html
@@ -1,5 +1,5 @@
-Function reference • dqrng Function reference • dqrng
@@ -17,7 +17,7 @@
@@ -106,7 +106,7 @@ All functions
diff --git a/reference/user-supplied-rng.html b/reference/user-supplied-rng.html
index cf5a95f..e7692dc 100644
--- a/reference/user-supplied-rng.html
+++ b/reference/user-supplied-rng.html
@@ -1,5 +1,5 @@
-Registering as user-supplied RNG — register_methods • dqrng Registering as user-supplied RNG — register_methods • dqrng
dqrng.Rmd
library(dqrng)
@@ -174,46 +174,46 @@ Usage from Rsystem.time(cat("pi ~= ", piR(N, rng = dqrng::dqrunif), "\n"))
#> pi ~= 3.141457
#> user system elapsed
-#> 0.222 0.041 0.262
system.time(stats::runif(N))
#> user system elapsed
-#> 0.078 0.013 0.090
+#> 0.087 0.003 0.090
system.time(dqrng::dqrunif(N))
#> user system elapsed
-#> 0.049 0.008 0.056
system.time(stats::rexp(N))
#> user system elapsed
-#> 0.338 0.004 0.342
+#> 0.342 0.005 0.346
system.time(dqrng::dqrexp(N))
#> user system elapsed
-#> 0.063 0.008 0.071
system.time(stats::rnorm(N))
#> user system elapsed
-#> 0.367 0.000 0.367
+#> 0.366 0.000 0.366
system.time(dqrng::dqrnorm(N))
#> user system elapsed
-#> 0.086 0.008 0.093
system.time(for (i in 1:100) sample.int(N, N/100, replace = TRUE))
#> user system elapsed
-#> 0.573 0.004 0.577
+#> 0.576 0.000 0.576
system.time(for (i in 1:100) dqsample.int(N, N/100, replace = TRUE))
#> user system elapsed
-#> 0.031 0.016 0.047
+#> 0.031 0.017 0.047
system.time(for (i in 1:100) sample.int(N, N/100))
#> user system elapsed
-#> 1.533 0.304 1.837
+#> 1.430 0.293 1.722
system.time(for (i in 1:100) dqsample.int(N, N/100))
#> user system elapsed
-#> 0.061 0.008 0.069set.seed() and
dqset.seed() influence both (dq)runif and
@@ -373,15 +373,14 @@ dqrnorm. Generally speaking you can use any
C++11 compliant RNG with 64 bit output size. For example, here the 64
@@ -395,16 +394,15 @@ dqrng_threefry.h
header file.dqrnorm, but the
technique is useful to support distributions not (yet) included in
dqrng. Note however, that the algorithms used for the distributions from
@@ -560,7 +557,7 @@ Site built with pkgdown 2.0.7.
+Site built with pkgdown 2.0.8.
diff --git a/articles/index.html b/articles/index.html index b8d0e76..26dfae2 100644 --- a/articles/index.html +++ b/articles/index.html @@ -1,5 +1,5 @@ -All vignettes
diff --git a/articles/parallel.html b/articles/parallel.html index 4111e95..795fb52 100644 --- a/articles/parallel.html +++ b/articles/parallel.html @@ -6,7 +6,7 @@Parallel RNG usage
Ralf Stubner
-2024-04-09
+2024-04-15
Source:vignettes/parallel.Rmd
parallel.RmdXo(ro)shiro: jump ahead with OpenMP// [[Rcpp::export]]
std::vector<double> parallel_random_sum(int n, int m, int ncores) {
dqrng::uniform_distribution dist(0.0, 1.0); // Uniform distribution [0,1)
- dqrng::xoshiro256plusplus rng(42); // properly seeded rng
+ dqrng::xoshiro256plusplus rng(42); // properly seeded rng
std::vector<double> res(m);
// ok to use rng here
@@ -239,6 +239,71 @@ Xo(ro)shiro: jump ahead with OpenMP
+In the previous example it was necessary to create our own RNG
+object. It would be nice if it were possible to use dqrng’s global RNG
+instead. To showcase this, we implement the same procedure as before,
+but this time make use of dqrng::random_64bit_accessor to
+access the global RNG and its clone(stream) method that
+creates a cloned version with a different stream:
+#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng.h>
+#include <dqrng_distribution.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::plugins(openmp)]]
+#include <omp.h>
+
+// [[Rcpp::export]]
+std::vector<double> random_sum(int n, int m) {
+ dqrng::uniform_distribution dist(0.0, 1.0); // Uniform distribution [0,1)
+ auto rng = dqrng::random_64bit_accessor{}; // properly seeded rng
+ std::vector<double> res(m);
+ for (int i = 0; i < m; ++i) {
+ double lres(0);
+ for (int j = 0; j < n; ++j) {
+ lres += dist(rng);
+ }
+ res[i] = lres / n;
+ }
+ return res;
+}
+
+// [[Rcpp::export]]
+std::vector<double> parallel_random_sum(int n, int m, int ncores) {
+ dqrng::uniform_distribution dist(0.0, 1.0); // Uniform distribution [0,1)
+ auto rng = dqrng::random_64bit_accessor{}; // properly seeded rng
+ std::vector<double> res(m);
+ // ok to use rng here
+
+#pragma omp parallel num_threads(ncores)
+{
+ // make thread local copy of rng and advance it by 1 ... ncores jumps
+ auto lrng = rng.clone(omp_get_thread_num() + 1);
+
+#pragma omp for
+ for (int i = 0; i < m; ++i) {
+ double lres(0);
+ for (int j = 0; j < n; ++j) {
+ lres += dist(*lrng);
+ }
+ res[i] = lres / n;
+ }
+}
+// ok to use rng here
+return res;
+}
+
+/*** R
+bm <- bench::mark(
+ parallel_random_sum(1e7, 8, 4),
+ random_sum(1e7, 8),
+ check = FALSE
+)
+bm[,1:4]
+*/
+Notes that the thread local RNG uses std::unique_ptr
+instead of Rcpp::Xptr since it is used in parallel
+code.
PCG: multiple streams with RcppParallel
@@ -260,66 +325,110 @@ PCG: multiple streams with RcppP
wrap it as dqrng::generator. The resulting correlation
matrix should be close to the identity matrix if the different streams
are independent:
-#include <Rcpp.h>
-// [[Rcpp::depends(dqrng, BH)]]
-#include <pcg_random.hpp>
-#include <dqrng_sample.h>
-// [[Rcpp::plugins(cpp11)]]
-// [[Rcpp::depends(RcppParallel)]]
-#include <RcppParallel.h>
-
-struct RandomFill : public RcppParallel::Worker {
- RcppParallel::RMatrix<int> output;
- uint64_t seed;
-
- RandomFill(Rcpp::IntegerMatrix output, const uint64_t seed) : output(output), seed(seed) {};
-
- void operator()(std::size_t begin, std::size_t end) {
- auto rng = dqrng::generator<pcg64>(seed, end);
- for (std::size_t col = begin; col < end; ++col) {
- auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*rng, 100000, output.nrow(), true);
- RcppParallel::RMatrix<int>::Column column = output.column(col);
- std::copy(sampled.begin(), sampled.end(), column.begin());
- }
- }
-};
-
-// [[Rcpp::export]]
-Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
- Rcpp::IntegerMatrix res(n, m);
- RandomFill randomFill(res, 42);
- RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
- return res;
-}
-
-/*** R
-res <- parallel_random_matrix(1e6, 8, 4)
-head(res)
-symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
- symbols = c(" ", "?", "!", "1"),
- abbr.colnames = FALSE, corr = TRUE)
-*/
+#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng_generator.h>
+#include <dqrng_sample.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+
+struct RandomFill : public RcppParallel::Worker {
+ RcppParallel::RMatrix<int> output;
+ uint64_t seed;
+
+ RandomFill(Rcpp::IntegerMatrix output, const uint64_t seed) : output(output), seed(seed) {};
+
+ void operator()(std::size_t begin, std::size_t end) {
+ auto rng = dqrng::generator<pcg64>(seed, end);
+ for (std::size_t col = begin; col < end; ++col) {
+ auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*rng, 100000, output.nrow(), true);
+ RcppParallel::RMatrix<int>::Column column = output.column(col);
+ std::copy(sampled.begin(), sampled.end(), column.begin());
+ }
+ }
+};
+
+// [[Rcpp::export]]
+Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
+ Rcpp::IntegerMatrix res(n, m);
+ RandomFill randomFill(res, 42);
+ RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
+ return res;
+}
+
+/*** R
+res <- parallel_random_matrix(1e6, 8, 4)
+head(res)
+symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
+ symbols = c(" ", "?", "!", "1"),
+ abbr.colnames = FALSE, corr = TRUE)
+*/
Head of the random matrix:
[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8]
-[1,] 67984 16279 69262 7126 21441 37720 51107 51045
-[2,] 69310 21713 82885 81157 54051 5261 91165 17833
-[3,] 76742 31232 78953 4626 94939 29416 85652 78296
-[4,] 76349 47427 1770 37957 33888 59134 94591 65793
-[5,] 85008 89224 43493 7925 60866 2464 14080 10763
-[6,] 38017 88509 51195 73086 1883 68193 75259 62216
+[1,] 16144 89512 16144 53103 16144 41760 16144 99788
+[2,] 21200 86552 21199 38327 21198 90293 21198 45713
+[3,] 124 60045 97553 79574 99977 84423 87126 35715
+[4,] 6590 93044 18536 98945 66969 53338 38426 53560
+[5,] 80618 42051 15815 79380 71590 98006 87220 63036
+[6,] 49954 90168 12614 32488 54145 68029 34733 48989
Correlation matrix:
[1,] 1
[2,] 1
-[3,] ? 1
-[4,] ? 1
-[5,] 1
-[6,] ? ? ? 1
-[7,] ? 1
-[8,] ? 1
+[3,] ? 1
+[4,] ? 1
+[5,] ? ? 1
+[6,] ? 1
+[7,] ? 1
+[8,] ? ? 1
attr(,"legend")
[1] 0 ‘ ’ 0.001 ‘?’ 0.003 ‘!’ 0.999 ‘1’ 1
So as expected the correlation matrix is almost equal to the identity
matrix.
+As in the previous section, we can again switch from seeding our own
+RNG to starting from the global RNG and selecting separate substreams
+from there. RNG selection and seeding can then be handled from R:
+#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng.h>
+#include <dqrng_sample.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+
+struct RandomFill : public RcppParallel::Worker {
+ RcppParallel::RMatrix<int> output;
+ dqrng::random_64bit_accessor rng{};
+
+ RandomFill(Rcpp::IntegerMatrix output) : output(output) {};
+
+ void operator()(std::size_t begin, std::size_t end) {
+ auto lrng = rng.clone(end);
+ for (std::size_t col = begin; col < end; ++col) {
+ auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*lrng, 100000, output.nrow(), true);
+ RcppParallel::RMatrix<int>::Column column = output.column(col);
+ std::copy(sampled.begin(), sampled.end(), column.begin());
+ }
+ }
+};
+
+// [[Rcpp::export]]
+Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
+ Rcpp::IntegerMatrix res(n, m);
+ RandomFill randomFill(res);
+ RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
+ return res;
+}
+
+/*** R
+dqRNGkind("pcg64")
+dqset.seed(42)
+res <- parallel_random_matrix(1e6, 8, 4)
+head(res)
+symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
+ symbols = c(" ", "?", "!", "1"),
+ abbr.colnames = FALSE, corr = TRUE)
+*/
@@ -346,7 +455,7 @@ PCG: multiple streams with RcppP
diff --git a/articles/sample.html b/articles/sample.html
index aa28311..25dbcc8 100644
--- a/articles/sample.html
+++ b/articles/sample.html
@@ -6,7 +6,7 @@
Fast sampling methods • dqrng
-
+
@@ -33,7 +33,7 @@
@@ -91,7 +91,7 @@ Fast sampling methods
Ralf
Stubner
- 2024-04-09
+ 2024-04-15
Source: vignettes/sample.Rmd
sample.Rmd
@@ -418,7 +418,7 @@ Technicalities
-Site built with pkgdown 2.0.7.
+Site built with pkgdown 2.0.8.
diff --git a/authors.html b/authors.html
index 5b680d0..41e3ef5 100644
--- a/authors.html
+++ b/authors.html
@@ -1,5 +1,5 @@
-Authors and Citation • dqrng Authors and Citation • dqrng
@@ -17,7 +17,7 @@
@@ -63,7 +63,7 @@
diff --git a/index.html b/index.html
index 3a92a7e..ed1225b 100644
--- a/index.html
+++ b/index.html
@@ -6,26 +6,13 @@
Fast Pseudo Random Number Generators • dqrng
-
+
-
+
Changelog • dqrng Changelog • dqrng
@@ -17,7 +17,7 @@
@@ -163,7 +163,7 @@ dqrng 0.0.1
diff --git a/pkgdown.yml b/pkgdown.yml
index 6e46253..eee6352 100644
--- a/pkgdown.yml
+++ b/pkgdown.yml
@@ -1,10 +1,10 @@
pandoc: 3.1.11
-pkgdown: 2.0.7
+pkgdown: 2.0.8
pkgdown_sha: ~
articles:
cpp-api: cpp-api.html
dqrng: dqrng.html
parallel: parallel.html
sample: sample.html
-last_built: 2024-04-09T20:28Z
+last_built: 2024-04-15T09:24Z
diff --git a/reference/dqrmv-1.png b/reference/dqrmv-1.png
index 64bdb36..2484a48 100644
Binary files a/reference/dqrmv-1.png and b/reference/dqrmv-1.png differ
diff --git a/reference/dqrmv.html b/reference/dqrmv.html
index a919916..709ac47 100644
--- a/reference/dqrmv.html
+++ b/reference/dqrmv.html
@@ -1,5 +1,5 @@
-Multivariate Distributions — dqrmvnorm • dqrng Multivariate Distributions — dqrmvnorm • dqrng
@@ -17,7 +17,7 @@
@@ -101,11 +101,11 @@ Examples
sigma <- matrix(c(4,2,2,3), ncol=2)
x <- dqrmvnorm(n=500, mean=c(1,2), sigma=sigma)
colMeans(x)
-#> [1] 0.9358251 2.0251444
+#> [1] 0.8614361 2.0263967
var(x)
#> [,1] [,2]
-#> [1,] 3.909685 1.884089
-#> [2,] 1.884089 2.976024
+#> [1,] 3.930309 1.917535
+#> [2,] 1.917535 2.881480
plot(x)
@@ -122,7 +122,7 @@ Examples
diff --git a/reference/dqrng-functions.html b/reference/dqrng-functions.html
index 7d471a1..29ebeff 100644
--- a/reference/dqrng-functions.html
+++ b/reference/dqrng-functions.html
@@ -1,5 +1,5 @@
-R interface — dqRNGkind • dqrng R interface — dqRNGkind • dqrng
dqrng
- 0.3.2.2
+ 0.3.2.4
@@ -250,7 +250,7 @@ Examples
diff --git a/reference/dqrng-package.html b/reference/dqrng-package.html
index e12ea13..99547af 100644
--- a/reference/dqrng-package.html
+++ b/reference/dqrng-package.html
@@ -1,5 +1,5 @@
-dqrng: Fast Pseudo Random Number Generators — dqrng-package • dqrng dqrng: Fast Pseudo Random Number Generators — dqrng-package • dqrng Unbiased Random Samples and Permutations — dqsample • dqrng Unbiased Random Samples and Permutations — dqsample • dqrng
@@ -17,7 +17,7 @@
@@ -116,7 +116,7 @@ See also
diff --git a/reference/generateSeedVectors.html b/reference/generateSeedVectors.html
index ebd8b72..b58534a 100644
--- a/reference/generateSeedVectors.html
+++ b/reference/generateSeedVectors.html
@@ -1,5 +1,5 @@
-Generate seed as a integer vector — generateSeedVectors • dqrng Generate seed as a integer vector — generateSeedVectors • dqrng
@@ -17,7 +17,7 @@
@@ -179,7 +179,7 @@ Examples
diff --git a/reference/index.html b/reference/index.html
index 6aedbb9..1d9898d 100644
--- a/reference/index.html
+++ b/reference/index.html
@@ -1,5 +1,5 @@
-Function reference • dqrng Function reference • dqrng
@@ -17,7 +17,7 @@
@@ -106,7 +106,7 @@ All functions
diff --git a/reference/user-supplied-rng.html b/reference/user-supplied-rng.html
index cf5a95f..e7692dc 100644
--- a/reference/user-supplied-rng.html
+++ b/reference/user-supplied-rng.html
@@ -1,5 +1,5 @@
-Registering as user-supplied RNG — register_methods • dqrng Registering as user-supplied RNG — register_methods • dqrng
In the previous example it was necessary to create our own RNG
+object. It would be nice if it were possible to use dqrng’s global RNG
+instead. To showcase this, we implement the same procedure as before,
+but this time make use of dqrng::random_64bit_accessor to
+access the global RNG and its clone(stream) method that
+creates a cloned version with a different stream:
#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng.h>
+#include <dqrng_distribution.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::plugins(openmp)]]
+#include <omp.h>
+
+// [[Rcpp::export]]
+std::vector<double> random_sum(int n, int m) {
+ dqrng::uniform_distribution dist(0.0, 1.0); // Uniform distribution [0,1)
+ auto rng = dqrng::random_64bit_accessor{}; // properly seeded rng
+ std::vector<double> res(m);
+ for (int i = 0; i < m; ++i) {
+ double lres(0);
+ for (int j = 0; j < n; ++j) {
+ lres += dist(rng);
+ }
+ res[i] = lres / n;
+ }
+ return res;
+}
+
+// [[Rcpp::export]]
+std::vector<double> parallel_random_sum(int n, int m, int ncores) {
+ dqrng::uniform_distribution dist(0.0, 1.0); // Uniform distribution [0,1)
+ auto rng = dqrng::random_64bit_accessor{}; // properly seeded rng
+ std::vector<double> res(m);
+ // ok to use rng here
+
+#pragma omp parallel num_threads(ncores)
+{
+ // make thread local copy of rng and advance it by 1 ... ncores jumps
+ auto lrng = rng.clone(omp_get_thread_num() + 1);
+
+#pragma omp for
+ for (int i = 0; i < m; ++i) {
+ double lres(0);
+ for (int j = 0; j < n; ++j) {
+ lres += dist(*lrng);
+ }
+ res[i] = lres / n;
+ }
+}
+// ok to use rng here
+return res;
+}
+
+/*** R
+bm <- bench::mark(
+ parallel_random_sum(1e7, 8, 4),
+ random_sum(1e7, 8),
+ check = FALSE
+)
+bm[,1:4]
+*/Notes that the thread local RNG uses std::unique_ptr
+instead of Rcpp::Xptr since it is used in parallel
+code.
PCG: multiple streams with RcppParallel
@@ -260,66 +325,110 @@ PCG: multiple streams with RcppP
wrap it as dqrng::generator. The resulting correlation
matrix should be close to the identity matrix if the different streams
are independent:
-#include <Rcpp.h>
-// [[Rcpp::depends(dqrng, BH)]]
-#include <pcg_random.hpp>
-#include <dqrng_sample.h>
-// [[Rcpp::plugins(cpp11)]]
-// [[Rcpp::depends(RcppParallel)]]
-#include <RcppParallel.h>
-
-struct RandomFill : public RcppParallel::Worker {
- RcppParallel::RMatrix<int> output;
- uint64_t seed;
-
- RandomFill(Rcpp::IntegerMatrix output, const uint64_t seed) : output(output), seed(seed) {};
-
- void operator()(std::size_t begin, std::size_t end) {
- auto rng = dqrng::generator<pcg64>(seed, end);
- for (std::size_t col = begin; col < end; ++col) {
- auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*rng, 100000, output.nrow(), true);
- RcppParallel::RMatrix<int>::Column column = output.column(col);
- std::copy(sampled.begin(), sampled.end(), column.begin());
- }
- }
-};
-
-// [[Rcpp::export]]
-Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
- Rcpp::IntegerMatrix res(n, m);
- RandomFill randomFill(res, 42);
- RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
- return res;
-}
-
-/*** R
-res <- parallel_random_matrix(1e6, 8, 4)
-head(res)
-symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
- symbols = c(" ", "?", "!", "1"),
- abbr.colnames = FALSE, corr = TRUE)
-*/
+#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng_generator.h>
+#include <dqrng_sample.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+
+struct RandomFill : public RcppParallel::Worker {
+ RcppParallel::RMatrix<int> output;
+ uint64_t seed;
+
+ RandomFill(Rcpp::IntegerMatrix output, const uint64_t seed) : output(output), seed(seed) {};
+
+ void operator()(std::size_t begin, std::size_t end) {
+ auto rng = dqrng::generator<pcg64>(seed, end);
+ for (std::size_t col = begin; col < end; ++col) {
+ auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*rng, 100000, output.nrow(), true);
+ RcppParallel::RMatrix<int>::Column column = output.column(col);
+ std::copy(sampled.begin(), sampled.end(), column.begin());
+ }
+ }
+};
+
+// [[Rcpp::export]]
+Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
+ Rcpp::IntegerMatrix res(n, m);
+ RandomFill randomFill(res, 42);
+ RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
+ return res;
+}
+
+/*** R
+res <- parallel_random_matrix(1e6, 8, 4)
+head(res)
+symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
+ symbols = c(" ", "?", "!", "1"),
+ abbr.colnames = FALSE, corr = TRUE)
+*/
dqrng::generator. The resulting correlation
matrix should be close to the identity matrix if the different streams
are independent:
-#include <Rcpp.h>
-// [[Rcpp::depends(dqrng, BH)]]
-#include <pcg_random.hpp>
-#include <dqrng_sample.h>
-// [[Rcpp::plugins(cpp11)]]
-// [[Rcpp::depends(RcppParallel)]]
-#include <RcppParallel.h>
-
-struct RandomFill : public RcppParallel::Worker {
- RcppParallel::RMatrix<int> output;
- uint64_t seed;
-
- RandomFill(Rcpp::IntegerMatrix output, const uint64_t seed) : output(output), seed(seed) {};
-
- void operator()(std::size_t begin, std::size_t end) {
- auto rng = dqrng::generator<pcg64>(seed, end);
- for (std::size_t col = begin; col < end; ++col) {
- auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*rng, 100000, output.nrow(), true);
- RcppParallel::RMatrix<int>::Column column = output.column(col);
- std::copy(sampled.begin(), sampled.end(), column.begin());
- }
- }
-};
-
-// [[Rcpp::export]]
-Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
- Rcpp::IntegerMatrix res(n, m);
- RandomFill randomFill(res, 42);
- RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
- return res;
-}
-
-/*** R
-res <- parallel_random_matrix(1e6, 8, 4)
-head(res)
-symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
- symbols = c(" ", "?", "!", "1"),
- abbr.colnames = FALSE, corr = TRUE)
-*/#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng_generator.h>
+#include <dqrng_sample.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+
+struct RandomFill : public RcppParallel::Worker {
+ RcppParallel::RMatrix<int> output;
+ uint64_t seed;
+
+ RandomFill(Rcpp::IntegerMatrix output, const uint64_t seed) : output(output), seed(seed) {};
+
+ void operator()(std::size_t begin, std::size_t end) {
+ auto rng = dqrng::generator<pcg64>(seed, end);
+ for (std::size_t col = begin; col < end; ++col) {
+ auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*rng, 100000, output.nrow(), true);
+ RcppParallel::RMatrix<int>::Column column = output.column(col);
+ std::copy(sampled.begin(), sampled.end(), column.begin());
+ }
+ }
+};
+
+// [[Rcpp::export]]
+Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
+ Rcpp::IntegerMatrix res(n, m);
+ RandomFill randomFill(res, 42);
+ RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
+ return res;
+}
+
+/*** R
+res <- parallel_random_matrix(1e6, 8, 4)
+head(res)
+symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
+ symbols = c(" ", "?", "!", "1"),
+ abbr.colnames = FALSE, corr = TRUE)
+*/Head of the random matrix:
[,1] [,2] [,3] [,4] [,5] [,6] [,7] [,8]
-[1,] 67984 16279 69262 7126 21441 37720 51107 51045
-[2,] 69310 21713 82885 81157 54051 5261 91165 17833
-[3,] 76742 31232 78953 4626 94939 29416 85652 78296
-[4,] 76349 47427 1770 37957 33888 59134 94591 65793
-[5,] 85008 89224 43493 7925 60866 2464 14080 10763
-[6,] 38017 88509 51195 73086 1883 68193 75259 62216Correlation matrix:
[1,] 1
[2,] 1
-[3,] ? 1
-[4,] ? 1
-[5,] 1
-[6,] ? ? ? 1
-[7,] ? 1
-[8,] ? 1
+[3,] ? 1
+[4,] ? 1
+[5,] ? ? 1
+[6,] ? 1
+[7,] ? 1
+[8,] ? ? 1
attr(,"legend")
[1] 0 ‘ ’ 0.001 ‘?’ 0.003 ‘!’ 0.999 ‘1’ 1So as expected the correlation matrix is almost equal to the identity matrix.
+As in the previous section, we can again switch from seeding our own +RNG to starting from the global RNG and selecting separate substreams +from there. RNG selection and seeding can then be handled from R:
+#include <Rcpp.h>
+// [[Rcpp::depends(dqrng, BH)]]
+#include <dqrng.h>
+#include <dqrng_sample.h>
+// [[Rcpp::plugins(cpp11)]]
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+
+struct RandomFill : public RcppParallel::Worker {
+ RcppParallel::RMatrix<int> output;
+ dqrng::random_64bit_accessor rng{};
+
+ RandomFill(Rcpp::IntegerMatrix output) : output(output) {};
+
+ void operator()(std::size_t begin, std::size_t end) {
+ auto lrng = rng.clone(end);
+ for (std::size_t col = begin; col < end; ++col) {
+ auto sampled = dqrng::sample::sample<std::vector<int>, uint32_t>(*lrng, 100000, output.nrow(), true);
+ RcppParallel::RMatrix<int>::Column column = output.column(col);
+ std::copy(sampled.begin(), sampled.end(), column.begin());
+ }
+ }
+};
+
+// [[Rcpp::export]]
+Rcpp::IntegerMatrix parallel_random_matrix(const int n, const int m, const int ncores) {
+ Rcpp::IntegerMatrix res(n, m);
+ RandomFill randomFill(res);
+ RcppParallel::parallelFor(0, m, randomFill, m/ncores + 1);
+ return res;
+}
+
+/*** R
+dqRNGkind("pcg64")
+dqset.seed(42)
+res <- parallel_random_matrix(1e6, 8, 4)
+head(res)
+symnum(x = cor(res), cutpoints = c(0.001, 0.003, 0.999),
+ symbols = c(" ", "?", "!", "1"),
+ abbr.colnames = FALSE, corr = TRUE)
+*/@@ -346,7 +455,7 @@
PCG: multiple streams with RcppP
diff --git a/articles/sample.html b/articles/sample.html
index aa28311..25dbcc8 100644
--- a/articles/sample.html
+++ b/articles/sample.html
@@ -6,7 +6,7 @@
Fast sampling methods • dqrng
-
+
@@ -33,7 +33,7 @@
Fast sampling methods
Ralf Stubner
-2024-04-09
+2024-04-15
Source:vignettes/sample.Rmd
sample.RmdTechnicalities
-Site built with pkgdown 2.0.7.
+Site built with pkgdown 2.0.8.
diff --git a/authors.html b/authors.html
index 5b680d0..41e3ef5 100644
--- a/authors.html
+++ b/authors.html
@@ -1,5 +1,5 @@
-Authors and Citation • dqrng Authors and Citation • dqrng
@@ -17,7 +17,7 @@
@@ -63,7 +63,7 @@
diff --git a/index.html b/index.html
index 3a92a7e..ed1225b 100644
--- a/index.html
+++ b/index.html
@@ -6,26 +6,13 @@
Fast Pseudo Random Number Generators • dqrng
-
+
-
+
Changelog • dqrng Changelog • dqrng
@@ -17,7 +17,7 @@
@@ -163,7 +163,7 @@ dqrng 0.0.1
diff --git a/pkgdown.yml b/pkgdown.yml
index 6e46253..eee6352 100644
--- a/pkgdown.yml
+++ b/pkgdown.yml
@@ -1,10 +1,10 @@
pandoc: 3.1.11
-pkgdown: 2.0.7
+pkgdown: 2.0.8
pkgdown_sha: ~
articles:
cpp-api: cpp-api.html
dqrng: dqrng.html
parallel: parallel.html
sample: sample.html
-last_built: 2024-04-09T20:28Z
+last_built: 2024-04-15T09:24Z
diff --git a/reference/dqrmv-1.png b/reference/dqrmv-1.png
index 64bdb36..2484a48 100644
Binary files a/reference/dqrmv-1.png and b/reference/dqrmv-1.png differ
diff --git a/reference/dqrmv.html b/reference/dqrmv.html
index a919916..709ac47 100644
--- a/reference/dqrmv.html
+++ b/reference/dqrmv.html
@@ -1,5 +1,5 @@
-Multivariate Distributions — dqrmvnorm • dqrng Multivariate Distributions — dqrmvnorm • dqrng
@@ -17,7 +17,7 @@
@@ -101,11 +101,11 @@ Examples
sigma <- matrix(c(4,2,2,3), ncol=2)
x <- dqrmvnorm(n=500, mean=c(1,2), sigma=sigma)
colMeans(x)
-#> [1] 0.9358251 2.0251444
+#> [1] 0.8614361 2.0263967
var(x)
#> [,1] [,2]
-#> [1,] 3.909685 1.884089
-#> [2,] 1.884089 2.976024
+#> [1,] 3.930309 1.917535
+#> [2,] 1.917535 2.881480
plot(x)
@@ -122,7 +122,7 @@ Examples
diff --git a/reference/dqrng-functions.html b/reference/dqrng-functions.html
index 7d471a1..29ebeff 100644
--- a/reference/dqrng-functions.html
+++ b/reference/dqrng-functions.html
@@ -1,5 +1,5 @@
-R interface — dqRNGkind • dqrng R interface — dqRNGkind • dqrng
dqrng
- 0.3.2.2
+ 0.3.2.4
@@ -250,7 +250,7 @@ Examples
diff --git a/reference/dqrng-package.html b/reference/dqrng-package.html
index e12ea13..99547af 100644
--- a/reference/dqrng-package.html
+++ b/reference/dqrng-package.html
@@ -1,5 +1,5 @@
-dqrng: Fast Pseudo Random Number Generators — dqrng-package • dqrng dqrng: Fast Pseudo Random Number Generators — dqrng-package • dqrng Unbiased Random Samples and Permutations — dqsample • dqrng Unbiased Random Samples and Permutations — dqsample • dqrng
@@ -17,7 +17,7 @@
@@ -116,7 +116,7 @@ See also
diff --git a/reference/generateSeedVectors.html b/reference/generateSeedVectors.html
index ebd8b72..b58534a 100644
--- a/reference/generateSeedVectors.html
+++ b/reference/generateSeedVectors.html
@@ -1,5 +1,5 @@
-Generate seed as a integer vector — generateSeedVectors • dqrng Generate seed as a integer vector — generateSeedVectors • dqrng
@@ -17,7 +17,7 @@
@@ -179,7 +179,7 @@ Examples
diff --git a/reference/index.html b/reference/index.html
index 6aedbb9..1d9898d 100644
--- a/reference/index.html
+++ b/reference/index.html
@@ -1,5 +1,5 @@
-Function reference • dqrng Function reference • dqrng
@@ -17,7 +17,7 @@
@@ -106,7 +106,7 @@ All functions
diff --git a/reference/user-supplied-rng.html b/reference/user-supplied-rng.html
index cf5a95f..e7692dc 100644
--- a/reference/user-supplied-rng.html
+++ b/reference/user-supplied-rng.html
@@ -1,5 +1,5 @@
-Registering as user-supplied RNG — register_methods • dqrng Registering as user-supplied RNG — register_methods • dqrng
Examples
sigma <- matrix(c(4,2,2,3), ncol=2)
x <- dqrmvnorm(n=500, mean=c(1,2), sigma=sigma)
colMeans(x)
-#> [1] 0.9358251 2.0251444
+#> [1] 0.8614361 2.0263967
var(x)
#> [,1] [,2]
-#> [1,] 3.909685 1.884089
-#> [2,] 1.884089 2.976024
+#> [1,] 3.930309 1.917535
+#> [2,] 1.917535 2.881480
plot(x)